an in silico skin absorption model for fragrance...

An in silico skin absorption model for fragrance materialsJie Shen a Lambros Kromidas a Terry Schultz b Sneha Bhatia a

a Research Institute for Fragrance Materials Inc 50 Tice Blvd Woodcliff Lake NJ 07677 USAb College of Veterinary Medicine The University of Tennessee 2407 River Dr Knoxville TN 37996 USA

A R T I C L E I N F O

Article historyReceived 6 August 2014Accepted 23 September 2014Available online 5 October 2014

KeywordsToxicologyQSARComputationalTopicalSafetyDermal

A B S T R A C T

Fragrance materials are widely used in cosmetics and other consumer products The Research Institutefor Fragrance Materials (RIFM) evaluates the safety of these ingredients and skin absorption is an im-portant parameter in refining systemic exposure Currently RIFMrsquos safety assessment process assumes100 skin absorption when experimental data are lacking This 100 absorption default is not support-able and alternate default values were proposed This study aims to develop and validate a practical skinabsorption model (SAM) specific for fragrance material It estimates skin absorption based on the meth-odology proposed by Kroes et al SAM uses three default absorption values based on the maximum flux(Jmax) ndash namely 10 40 and 80 Jmax may be calculated by using QSAR models that determine octanolwater partition coefficient (Kow) water solubility (S) and permeability coefficient (Kp) Each of these QSARmodels was refined and a semi-quantitative mechanistic model workflow is presented SAM was vali-dated with a large fragrance-focused data set containing 131 materials All resulted in predicted valuesfitting the three-tiered absorption scenario based on Jmax ranges This conservative SAM may be appliedwhen fragrance material lack skin absorption data

copy 2014 Elsevier Ltd All rights reserved

1 Introduction

Skin absorption is a very important parameter for safety assess-ment especially for topically applied fragrance materials Chemicalsin contact with the skin have the potential to be absorbed into theskin and enter the systemic circulation To enter the systemic cir-culation the chemical must reach the underlying dermis replete withcapillaries Skin absorption occurs by passive diffusion through theepidermis and directly by sweat glands and hair follicles (Ngo et al2010) Chemicals that penetrate no further than the epidermis aredestined to be eliminated by desquamation and thus not reach thesystemic circulation (Sundberg et al 2012) Therefore determin-ing the penetration of a substance is crucial for assessing systemicexposure

Usually skin absorption of a target material is obtained exper-imentally in vitro andor in vivo based on different species includingpig monkey and human subjects In the absence of experimentalskin absorption data it is customary by safety assessors to extrap-olate via read-across structural analogs or use default values TheEuropean Commission (EC) guidance on dermal absorption pro-posed two default values ndash 100 or 10 if the substance of interestis very lipophilic or very hydrophilic (ie log Kow ltminus1 or gt+4) witha MW gt500 (European Commission 2004) For pesticides the Eu-ropean Food Safety Authority (EFSA) replaced these EC default valueswith 25 for liquid concentrates and 75 when diluting them forspraying (EFSA Panel on Plant Protection Products and their Residues2011 2012) More recently however Aggarwal et al (2014) ana-lyzed human skin absorption data on pesticides that were availableuntil 2012 and proposed a 6 default value for liquid concentratesand 30 for spray dilutions

Currently in RIFMrsquos (Research Institute for Fragrance Materi-als) safety assessment process a 100 default absorption value isapplied for materials without experimental data (Belsito et al 20072011 2012a 2012b 2012c) However as discussed by Kroes et al(2007) ldquothe assumption of 100 absorption is not scientifically sup-portablerdquo and based on their analysis of 15 cosmetic ingredientsand 62 chemicals in the EDETOX database (Williams 2004) theyproposed three default skin absorption values for cosmetic ingre-dients Based on their derivation and analysis the three differentdefault skin absorption values proposed were based on the maximumflux (Jmax in unit of μgcm2h) Jmax is the theoretically achieved dosebased on Fickrsquos first law of diffusion (Fick 1855) when a material

Abbreviations RIFM Research Institute for Fragrance Materials SAM skin ab-sorption model QSAR Quantitative StructurendashActivity Relationship EC EuropeanCommission EFSA European Food Safety Authority EDETOX Evaluations and Pre-dictions of Dermal Absorption of Toxic Chemicals USEPA United States EnvironmentalProtection Agency ACTOR Aggregated Computational Toxicology Resource SE stan-dard error OECD Organisation for Economic Co-operation and Development RMSEroot-mean-square deviation TTC Threshold of Toxicological Concern MOE marginof exposure MOS margin of safety CAESAR Computer Assisted Evaluation of in-dustry chemical Substances According to Regulations

Corresponding author Research Institute for Fragrance Materials Inc 50 TiceBlvd Woodcliff Lake NJ 07677 USA Tel +1 201 689 8089 ext 130 fax 1-201-689-8090

E-mail address lkromidasrifmorg (L Kromidas)

httpdxdoiorg101016jfct2014090150278-6915copy 2014 Elsevier Ltd All rights reserved

Food and Chemical Toxicology 74 (2014) 164ndash176

Contents lists available at ScienceDirect

Food and Chemical Toxicology

journal homepage wwwelseviercom locate foodchemtox

is maintained in a saturated solution or at steady state equilibri-um whose flux describes the amount of permeant per unit time andarea (ie μgcm2h) (Magnusson et al 2004) Jmax is independentof the formulation in which the material contacts the skin and is aconstant value when the formulation does not change the skin barrier(Kroes et al 2007) Based on Jmax the default absorption valuesproposed were as follows

bull Material with Jmax le01 μgcm2h should be assigned a skin ab-sorption default value of less than 10

bull If the Jmax value is gt01 μgcm2h but le10 μgcm2h the defaultskin absorption assigned did not exceed 40

bull If a material had a Jmax of gt10 μgcm2h the default skin absorp-tion assigned was no more than 80

The three default skin absorption percentages were proposedto represent low absorbed medium absorbed and high absorbedmaterial These default values were derived from a broad rangeof 15 cosmetic ingredients and considered several worst-case as-sumptions such as (i) cosmetic ingredients are present at saturationlevels (ii) no depletion of the ingredient occurs during the expo-sure period (iii) the formulation does not affect the skin barrierand (iv) by using the maximal flux over the entire exposure timethe lower flux during the lag time is ignored (Kroes et al 2007)In another study Guy proved that the skin absorption of chemi-cals may be classified by their calculated Jmax (Guy 2010) Usingthe same model the Jmax of 20 fragrance materials were calculatedand 16 of them were taken one step further to calculate their ab-sorption percentage by including the applied dose area and timeOver-prediction was observed in 14 materials and thereforethis approach could also be considered an extreme estimation ofabsorption

Based on the abovementioned studies it is apparent that the keyto determining a substancersquos default absorption value is to get anaccurate Jmax Centered on this reasoning we propose an in silico semi-quantitative mechanistic model for assigning the same default skinabsorption values as proposed by Kroes et al (2007) but specifi-

cally constructed around the fragrance material (Fig 1) As we willshow their overall methodology can be applied to derive similarabsorption values for RIFMrsquos fragrance materials that lack experi-mental skin absorption data provided we tailor the model tospecifically fit a defined set of fragrance material physicochemicalparameters Our mechanistic model was validated with a fragrance-focused data set containing 131 materials All resulted in predictedvalues fitting the proposed three-tiered SAM based on Jmax ranges

2 Methodology development and data sets

21 Defining the chemical space of fragrance materials

Getting insight into the chemical space of fragrance materialsand where this space is located in the chemical universe was in-structive for us to get the big picture and develop a fragrance-focused SAM Herein 2120 fragrance materials with known chemicalstructure in RIFM database were reviewed The fragrance chemi-cal space was profiled using three physicochemical properties thatsignificantly influence the overall absorption of a topically appliedsubstance namely molecular weight (MW) log Kow (octanolwater partition coefficient) and water solubility (S) For perspectivethese were compared to more than half a million industry chemi-cals from the United States Environmental Protection Agency (USEPA)ACTOR database (Judson et al 2012)

As shown in Fig 2 the MW log Kow and log S values of ~500000industry chemicals from USEPA ACTOR database are calculatedusing EPI Suite (USEPA) and plotted to represent the chemicaluniverse (Fig 2A blue dots) The same parameters of 2120fragrance materials are also calculated and plotted in the samechart (Fig 2A red dots) Clearly not only the number of fragrancematerial is significantly small but the fragrance space is limitedFurther analysis indicates that more than 99 of fragrance mate-rials gave a MW ranging from 30 to 330 (Fig 2B) a log Kow fromminus1 to 9 (Fig 2C) and log S from minus9 to 1 (Fig 2D) As such we con-sider any materials falling within these ranges as ldquofragrance-likerdquomaterials

Fig 1 Workflow for applying the skin absorption model (SAM) in safety assessment For a target material first look for available skin absorption data If an experimentalvalue is not available look for experimentally derived Kp and water solubility values to calculate Jmax If an experimentally derived Kp is not available look for an experi-mentally derived log Kow or use consensus value to estimate the log Kow to determine a predicted Kp using Eq (4) This predicted Kp may be used with experimentallyderived or predicted solubility values to calculate Jmax Then the percent skin absorption is estimated based on Eq (8) If the fragrance material is an ester oneneeds to calculate the Jmax of the parent and breakdown products (ie carboxylic acid and alcohol moieties) For conservative purposes accept the value that gives the highestskinabsorption

165J Shen et alFood and Chemical Toxicology 74 (2014) 164ndash176

22 Datasets for validating and refining SAM

An important step in calculating Jmax is to get the Kp ndash the perme-ability coefficient Potts and Guyrsquos equation was proposed to be usedfor calculating the Kp (Potts and Guy 1992) In order to evaluate thesuitability of the Potts and Guyrsquos equation 105 materials falling withinthe fragrance chemical space with experimentally determined Kp andlog Kow values were collected as data set I Some of that data came fromthe Flynn data set (Flynn 1990) that was used to develop the originalKp equation by Potts and Guy (1992) The rest came from EDETOXdataset (Table 1) To validate SAM 131 materials falling within the fra-grance chemical space with experimentally determined (via in vitroor in vivo methods) percent skin absorption from using either humanor pig skin were collected as data set II (Table 2) Rat data were ex-cluded from this study due to the significant species difference betweenrat and human (van Ravenzwaay and Leibold 2004) Of the valida-tion material 54 were fragrance materials and 77 were ldquofragrance-likerdquo materials Forty-five (45) were collected from RIFMrsquos database (Apiet al 2013 Barber et al 1992 Bickers et al 2005 Bobin et al 1997Bronaugh et al 1990 Cross et al 1998 Ford et al 1999 2001 Gilpinet al 2010 Green et al 2008 Green et al 2014 Green and Brain 20012005 Green and Walters 2006 Hawkins et al 1994 1995 2002Hotchkiss 1998 Isola and Api 2002 Jimbo 1983 Keith 2003 Kraelingand Bronaugh 1997 2003 Liu and Hotchkiss 1997 Madsen et al 2011Politano et al 2013 Smith et al 2000 Tonge 1995 Watkinson et al1992 Yang et al 1994 Yano et al 1986) and 86 from the EDETOX da-tabase (Williams 2004) Of the 54 fragrance materials 52 had humandata (46 in vitro six in vivo) and two had pig data (in vitro) Of the 77

ldquofragrance-likerdquo materials 72 had human data (46 in vitro 26 in vivo)and five had pig data (two in vitro three in vivo) Our validation dataset contained 96 materials with experimentally determined log Kow81 with experimentally determined water solubility (Cwater

Sat) and 27had an experimentally determined Kp Twenty-two had all three pa-rameters experimentally derived

23 Calculating Jmax

As shown in Equation (Eq (1)) below Jmax is calculated from twoparameters ndash the permeability coefficient (Kp) used to character-ize the diffusion of chemicals through membranes expressed as cmh(Hostynek 2008) and saturated water solubility (Cwater

Sat in the unitof mgL or μgcm3)

J g cm h K cm h C g cmmax p watersatμ μ2 3( ) ( ) ( )= times (1)

231 Calculate Kp

As shown above one of the important parameters to calculateJmax is Kp In most cases the experimental determined Kp is notavailable Therefore a proper QSAR model could be useful to getan estimated Kp Kroes et al (2007) proposed to use Potts and Guyrsquosequation (Eq (2)) which was derived from the Flynn data set of93 materials with experimentally determined Kp (Flynn 1990)

log logK cm h K MWp ow( ) = minus + times minus times2 7 0 71 0 0061 (2)

(n = 93 r2 = 067 SE = 0794 F = 8405)

Fig 2 The fragrance chemical space depicted using MW log Kow and log S (A) The overlap of 2120 fragrance materials (red dots) from RIFMrsquos database on ~500000 in-dustry chemicals from USEPA ACTOR database (blue dots) (B) The histogram of MW of 2120 fragrance materials (C) The histogram of log Kow of 2120 fragrance materials(D) The histogram of log S of 2120 fragrance materials (For interpretation of the references to color in this figure legend the reader is referred to the web version of thisarticle)

166 J Shen et alFood and Chemical Toxicology 74 (2014) 164ndash176

Table 1Predicted and experimentally determined Kp and Kow of 105 fragrance and ldquofragrance-likerdquo materials

Material name CAS MW log Kow log Kp (cmh) Source

Exp Est Exp Guyrsquos Eq RIFMrsquos Eq

17α-Hydroxyprogesterone 68-96-2 33047 317 315 minus322 minus247 minus344 Flynn set2-(2-Butoxyethoxy)ethanol 112-34-5 16223 056 065 minus445 minus329 minus344 EDETOX2-(2-Ethoxyethoxy)ethanol 111-90-0 13418 minus054 minus032 minus388 minus390 minus394 EDETOX23-Butanediol 513-85-9 9012 minus092 minus041 minus440 minus390 minus371 Flynn set246-Trichlorophenol 25167-82-2 19745 369 357 minus123 minus128 minus148 Flynn set24-Dichlorophenol 120-83-2 16300 306 293 minus122 minus152 minus155 Flynn set2-Butanone 78-93-3 7211 029 052 minus235 minus293 minus258 Flynn set2-Butoxyethanol 111-76-2 11818 083 078 minus367 minus283 minus271 EDETOX2-Chlorophenol 25167-80-0 12856 215 225 minus148 minus196 minus183 Flynn set2-Cresol 95-48-7 10814 195 205 minus180 minus198 minus175 Flynn set2-Ethoxy ethanol 110-80-5 9012 minus032 minus023 minus360 minus348 minus326 Flynn set2-Isopropoxyethanol 109-59-1 10415 005 017 minus336 minus330 minus314 EDETOX2-Methoxyethanol 109-86-4 7610 minus077 minus064 minus254 minus371 minus343 EDETOX2-Naphthol 135-19-3 14417 270 271 minus155 minus166 minus160 Flynn set2-Phenoxyethanol 122-99-6 13817 116 118 minus287 minus272 minus270 EDETOX2-Phenylphenol 90-43-7 17021 309 321 minus180 minus154 minus161 EDETOX34-Xylenol 95-65-8 12217 223 252 minus144 minus186 minus170 Flynn set3-Cresol 108-39-4 10814 196 206 minus182 minus197 minus174 Flynn set3-Nitrophenol 554-84-7 13911 200 168 minus225 minus213 minus207 Flynn set3-Phenyl-1-propanol 1335-12-2 13620 188 196 minus128 minus220 minus213 EDETOX4-Bromophenol 106-41-2 17301 259 241 minus144 minus192 minus203 Flynn set4-Chlorocresol 1570-64-5 14259 278 272 minus126 minus160 minus152 Flynn set4-Chlorophenol 106-48-9 12856 239 227 minus144 minus179 minus165 Flynn set4-Cresol 1319-77-3 10814 194 206 minus175 minus198 minus175 Flynn set4-Ethyl phenol 123-07-9 12217 258 256 minus146 minus161 minus143 Flynn set4-Nitrophenol 100-02-7 13911 191 167 minus225 minus219 minus214 Flynn set5-Fluorouracil 51-21-8 13008 minus089 minus082 minus478 minus413 minus416 EDETOXAcetylsalicylic acid 50-78-2 18016 119 126 minus214 minus295 minus317 EDETOXAmobarbital 57-43-2 22628 207 192 minus264 minus261 minus305 Flynn setAtropine 51-55-8 28938 183 185 minus412 minus317 minus397 Flynn setBarbital 57-44-3 18420 065 067 minus395 minus336 minus363 Flynn setBenzene 71-43-2 7811 213 197 minus095 minus166 minus125 EDETOXBenzoic acid 65-85-0 12212 187 169 minus160 minus212 minus197 EDETOXBenzyl alcohol 100-51-6 10814 110 114 minus222 minus258 minus239 Flynn setBenzyl nicotinate 94-44-0 21324 240 234 minus180 minus230 minus264 EDETOXBoric acid 10043-35-3 6183 017 minus074 minus330 minus296 minus255 EDETOXButobarbital 77-28-1 21225 173 165 minus371 minus277 minus314 Flynn setButyl 4-hydroxybenzoate 94-26-8 19423 357 315 minus215 minus135 minus153 EDETOXButyl nicotinate 6938-06-3 17922 227 208 minus178 minus218 minus234 EDETOXButyric acid 107-92-6 8811 079 088 minus300 minus268 minus239 Flynn setCaffeine 58-08-2 19419 minus007 minus026 minus359 minus393 minus429 EDETOXChlorocresol 59-50-7 14259 310 278 minus126 minus137 minus128 EDETOXChloroxylenol 88-04-0 15661 327 322 minus128 minus133 minus132 Flynn setChlorpheniramine 113-92-8 27480 338 364 minus266 minus198 minus263 Flynn setcis-13-Dichloropropene 10061-01-5 11097 203 191 minus010 minus194 minus172 EDETOXCodeine 76-57-3 29937 119 141 minus431 minus368 minus458 Flynn setCortexone 64-85-7 33047 288 315 minus335 minus267 minus366 Flynn setCoumarin 91-64-5 14615 139 176 minus204 minus260 minus262 EDETOXDiclofenac 15307-86-5 29615 451 439 minus300 minus130 minus202 EDETOXDimethylethylamine 598-56-1 7314 070 062 minus240 minus265 minus228 EDETOXDimethylformamide 68-12-2 7310 minus101 minus060 minus202 minus386 minus358 EDETOXDMP 131-11-3 19419 160 175 minus448 minus275 minus303 EDETOXEphedrine 299-42-3 16524 113 110 minus222 minus291 minus304 Flynn setEstradiol 50-28-2 27239 401 382 minus240 minus151 minus212 Flynn setEstriol 50-27-1 28839 245 271 minus440 minus272 minus349 Flynn setEstrone 53-16-7 27037 313 385 minus244 minus213 minus276 Flynn setEthanol 64-17-5 4607 minus031 minus017 minus310 minus320 minus273 Flynn setEthyl benzene 100-41-4 10617 315 297 008 minus111 minus081 Flynn setEthyl ether 60-29-7 7412 089 090 minus180 minus252 minus215 Flynn setEthyl nicotinate 614-18-6 15117 132 110 minus218 minus268 minus273 EDETOXHeptanoic acid 111-14-8 13019 242 236 minus170 minus178 minus165 Flynn setHexanoic acid 142-62-1 11616 192 187 minus185 minus205 minus186 Flynn setIbuprofen 15687-27-1 20629 397 369 minus144 minus114 minus137 EDETOXIsoquinoline 119-65-3 12916 208 191 minus178 minus201 minus189 Flynn setm-Cresol 108-39-4 10814 196 206 minus182 minus197 minus174 EDETOXMDA 101-77-9 19827 159 221 minus264 minus278 minus308 EDETOXMeperidine 57-42-1 24734 272 261 minus243 minus228 minus280 Flynn setMethanol 67-56-1 3204 minus077 minus064 minus330 minus344 minus291 Flynn setMethyl nicotinate 93-60-7 13714 083 066 minus247 minus295 minus294 EDETOXMethyl-4-hydroxy benzoate 99-76-3 15215 196 172 minus204 minus224 minus226 Flynn setMorphine 57-27-2 28535 089 101 minus503 minus381 minus464 Flynn setNN-Diethyl-m-toluamide 134-62-3 19128 218 245 minus289 minus232 minus255 EDETOXNaproxen 22204-53-1 23027 318 307 minus254 minus185 minus225 Flynn set

(continued on next page)


Before applying Potts and Guyrsquos equation (Potts and Guy 1992)we refined the parameters by re-running the multiple linear re-gression based on the data set I As mentioned above data set Icontains 105 materials falling within the fragrance chemical spacewith experimentally determined log Kow and Kp Specifically the in-dependent variables are log Kow and MW the dependent variableis the log Kp Least square was performed to fit the parameters ofthe linear equation The updated Eq (3) with newly fitted param-eters was used to calculate log Kp for the materials falling in thefragrance chemical space


(n = 105 r2 = 0703 SE = 0594 F = 12056)Note that the standard error of estimation (SE) of the above equa-

tion (Eq (3)) is 0594 For conservative purposes this number isadded to the predicted log Kp before using it to calculate the JmaxTherefore the following Eq (4) is used to estimate the conserva-tive Kp


It is worth noting that the viable skin layers beneath the stratumcorneum could contribute to the penetration rate for very lipo-philic chemicals Cleek and Bunge (1993) took this into considerationand proposed the algebraic Eq (5) to modify for Kp Taking this intoconsideration Eq (1) is more correctly stated as Eq (6)

K cm hK

K MWpcorr p

p

( ) =+

times1

2 6

(5)

J g cm h K cm h C g cmmax pcorr

watersatμ μ2 3( ) ( ) ( )= times (6)

232 Calculate log Kow

As shown in Eq (4) the Kp is calculated based on log Kow andMW If the experimental log Kow is unknown different QSAR pre-diction models could be used to calculate the log Kow There aretwo key factors for a QSAR model molecular descriptors and train-ing algorithms Molecular descriptors are a series of parameterscalculated based on molecular structure which can be under-stood by a mathematics algorithm The training algorithm is thecore of building a QSAR model Different molecular descriptors andalgorithms could build QSAR models with different preferences andperformance It has been suggested that using a consensus modelbased on the average prediction of different models could yieldmore accurate results than individual models (Mannhold et al 2009Tetko et al 2009) Therefore in this work seven different modelswith different algorithms and descriptors are used to compose aconsensus model (Table 3) The seven models were selectedbased on molecular descriptor modeling algorithm training setand accessibility The idea is to cover different models with diversedescriptors algorithms and training sets to obtain a more repre-sentative outcome These seven models were also validated intheir original papers with a high accuracy The average value ofthe seven predictions is used as the final output of the predictedlog Kow

233 Calculate water solubility (CwaterSat)

As shown in Eq (1) another parameter for calculating Jmax

is the saturated water solubility (Cwatersat) in the units of mgL which

equals μgcm3 Like the log Kow when experimental data is lackingthe consensus prediction from four different QSAR models isused to estimate the water solubility (S) in the units of molL(Table 4) However one should note that the water solubility modelsare not as accurate as those of log Kow The root-mean-square

Table 1 (continued)



n-Butanol 71-36-3 7412 088 086 minus260 minus253 minus216 Flynn setn-Decanol 112-30-1 15829 457 390 minus110 minus042 minus035 Flynn setn-Heptanol 111-70-6 11621 262 237 minus150 minus155 minus133 Flynn setn-Hexanol 111-27-3 10218 203 187 minus189 minus188 minus161 Flynn setN-Hexyl nicotinate 23597-82-2 20727 351 306 minus175 minus147 minus173 EDETOXNicotine 54-11-5 16224 117 114 minus248 minus286 minus298 Flynn setNicotinic acid 59-67-6 12311 036 022 minus462 minus320 minus313 EDETOXNitroglycerine 100292-13-5 22709 162 161 minus196 minus294 minus340 Flynn setn-Nonanol 143-08-8 14426 377 338 minus122 minus090 minus079 Flynn setn-Octanol 111-87-5 13023 300 288 minus128 minus136 minus121 Flynn setn-Pentanol 71-41-0 8815 151 138 minus222 minus217 minus184 Flynn setn-Propanol 71-23-8 6010 025 035 minus285 minus289 minus247 Flynn seto-Cresol 95-48-7 10814 195 205 minus180 minus198 minus175 EDETOXOctanoic acid 124-07-2 14422 305 286 minus160 minus141 minus134 Flynn setParathion 56-38-2 29126 383 343 minus372 minus176 minus248 EDETOXp-Cresol 106-44-5 10814 194 206 minus176 minus198 minus175 EDETOXPentanoic acid 109-52-4 10213 139 138 minus270 minus234 minus210 Flynn setPhenobarbital 11097-06-6 23224 147 135 minus334 minus307 minus357 Flynn setPhenol 108-95-2 9411 146 155 minus209 minus224 minus195 Flynn setPregnenolone 145-13-1 31649 422 380 minus282 minus163 minus248 Flynn setProgesterone 57-83-0 31447 387 387 minus282 minus187 minus272 Flynn setPropoxur 114-26-1 20925 152 204 minus259 minus290 minus326 EDETOXPropranolol 525-66-6 25935 348 280 minus277 minus181 minus237 EDETOXResorcinol 108-46-3 11011 080 113 minus362 minus280 minus264 Flynn setSalicylic acid 69-72-7 13812 226 191 minus186 minus194 minus186 Flynn setScopolamine 51-34-3 30336 098 101 minus430 minus385 minus478 Flynn setStyrene 100-42-5 10415 295 272 minus019 minus124 minus094 Flynn setTestosterone 58-22-0 28843 332 329 minus266 minus210 minus283 Flynn setThymol 89-83-8 15022 330 332 minus128 minus127 minus122 Flynn setToluene 108-88-3 9214 273 248 000 minus132 minus097 Flynn setTrichloromethane 67-66-3 11938 197 184 minus180 minus203 minus186 EDETOXTrimethylamine 75-50-3 5911 016 017 minus372 minus295 minus253 EDETOX

exp = experimentally determined est = estimated


Table 2Estimated and experimentally determined skin absorption values for 131 fragrance and fragrance like materials

ID Material name CAS MW log Kow CwaterSat

(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax

(μgcm2h)Absorption ()

Est Exp Source

1 Musk ketone 81-14-1 29431 430a 19a minus1559 276E-02 233E-02 004 10 05 RIFM Db2 Acetyl cedrene 32388-55-9 2464 420 452 minus1071 850E-02 562E-02 254 40 1352 RIFM Db3 1-(12345678-Octahydro-2388-

tetramethyl-2-naphthalenyl)ethanone

54464-57-2 23438 431 60 minus0846 143E-01 775E-02 465 40 1651 RIFM Db

4 6-Acetyl-112447-hexamethyltetraline

21145-77-7 25841 570a 13a minus0073 845E-01 136E-01 017 40 088 EDETOX

5 Farnesol 4602-84-0 22237 475 276 minus0366 431E-01 124E-01 342 40 398 RIFM Db6 134678-Hexahydro-466788-

hexamethylcyclopenta-γ-2-benzopyran

1222-05-5 25841 590a 18a 0079 120E+00 143E-01 025 40 516 RIFM Db

7 α-Hexylcinnamaldehyde 101-86-0 21633 468 253 minus0348 449E-01 127E-01 321 40 954 RIFM Db8 p-t-Butyl-α-methylhydrocinnamic

aldehyde80-54-6 20431 407 427 minus0671 213E-01 982E-02 419 40 30 RIFM Db

9 Benzyl benzoate 120-51-4 21225 397a 1671 minus0841 144E-01 797E-02 1332 80 712 RIFM DbBenzoic acid 12212 187a 3400a minus1602a 250E-02 226E-02 7684Benzyl alcohol 10814 110a 42900a minus2220a 603E-03 588E-03 25241

10 Benzyl salicylate 118-58-1 22825 374 3578 minus1209 618E-02 454E-02 1626 80 15 RIFM Db2-Hydroxybenzoic acid 13812 226a 2240a minus1860a 138E-02 130E-02 2910Benzyl alcohol 10814 110a 42900a minus2220a 603E-03 588E-03 25241

11 Amyl salicylate 2050-08-0 20826 400 4211 minus0774 168E-01 870E-02 3662 80 103 RIFM Db2-Hydroxybenzoic acid 13812 226a 2240a minus1860a 138E-02 130E-02 2910Amyl alcohol 8815 151a 22000a minus2220a 603E-03 590E-03 12974

12 d-Limonene 5989-27-5 13624 438a 138a 0361 230E+00 203E-01 280 40 5 RIFM Db13 Butyl salicylate 2052-14-4 19423 463a 13269 minus0134 735E-01 149E-01 19743 80 171 RIFM Db

2-Hydroxybenzoic acid 13812 226a 2240a minus1860a 138E-02 130E-02 2910Butyl alcohol 7412 088a 63200a minus2600a 251E-03 249E-03 15744

14 α-Methyl-13-benzodioxole-5-propionaldehyde

1205-17-0 19222 214 1390 minus1996 101E-02 957E-03 1330 80 501 RIFM Db

15 Methyl dihydrojasmonate 24851-98-7 22632 280 280a minus1899 126E-02 118E-02 329 80 459 RIFM Db(3-oxo-2-pentylcyclopentyl)acetic acid 21229 259 5547 minus1896 127E-02 119E-02 6576Methanol 3204 minus077a 1000000a minus3300a 501E-04 501E-04 50064

16 Safrole 94-59-7 16219 280 5922 minus1142 721E-02 533E-02 3156 80 384 EDETOX17 Isoeugenol 97-54-1 16421 304a 1164 minus0986 103E-01 684E-02 7963 80 384 RIFM Db18 2-Methoxy-4-propylphenol 2785-87-7 16622 294 1531 minus1085 821E-02 584E-02 8936 80 226 RIFM Db19 2-Methoxy-4-vinylphenol 7786-61-0 15018 220 25852 minus1455 351E-02 301E-02 7785 80 384 RIFM Db20 dl-Citronellol 106-22-9 15627 391a 575 minus0232 586E-01 153E-01 8826 80 47 RIFM Db21 Eugenyl methyl ether 93-15-2 17823 294 500a minus1231 587E-02 451E-02 2256 80 497 RIFM Db22 Dihydromyrcenol 18479-58-8 15627 311 11772 minus0839 145E-01 854E-02 10049 80 567 RIFM Db23 Coumarin 91-64-5 14615 139a 1900a minus2040a 912E-03 875E-03 1662 80 597 RIFM Db24 Dihydro-α-terpineol 498-81-7 15627 297 16563 minus0943 114E-01 737E-02 12206 80 35 RIFM Db25 Benzyl acetate 140-11-4 15018 196a 3100a minus1640 229E-02 207E-02 6403 80 787 RIFM Db

Acetic acid 6005 minus017a 1000000a minus2194 640E-03 628E-03 628304Benzyl alcohol 10814 110a 42900a minus2220a 603E-03 588E-03 25241

26 Eugenol 97-53-0 16421 227a 2460a minus1571 269E-02 237E-02 5837 80 226 RIFM Db27 Cinnamic acid 621-82-9 14816 213a 570a minus1488 325E-02 282E-02 1609 80 608 RIFM Db28 Diethyl malonate 105-53-3 16017 096a 23200a minus2517 304E-03 299E-03 6946 80 543 EDETOX

Malonic acid 10406 minus081a 763000a minus3199 633E-04 631E-04 48167Ethyl alcohol 4607 minus031a 1000000a minus3100a 794E-04 793E-04 79268Ethyl hydrogen malonate 13212 014 2801413 minus2807 156E-03 155E-03 43421

29 Benzoic acid 65-85-0 12212 187a 3400a minus1602a 250E-02 226E-02 7685 80 606 EDETOX30 Methyl salicylate 119-36-8 15215 255a 700a minus1216 608E-02 472E-02 3304 80 307 RIFM Db

2-Hydroxybenzoic acid 13812 226a 2240a minus1860a 138E-02 130E-02 2910Methanol 3204 minus077a 1000000a minus3300a 501E-04 501E-04 50064

31 Geraniol 106-24-1 15425 347a 531a minus0542 287E-01 121E-01 6426 80 73 RIFM Db32 Linalool 78-70-6 15425 297a 1590a minus0922 120E-01 762E-02 12108 80 144 RIFM Db33 2-Hydroxybenzoic acid 69-72-7 13812 226a 2240a minus1860a 138E-02 130E-02 2910 80 2278 EDETOX34 Cinnamyl alcohol 104-54-1 13418 195a 71127 minus1459 347E-02 301E-02 21393 80 659 RIFM Db35 2-Phenoxyethanol 122-99-6 13817 116a 26700a minus2870a 135E-03 134E-03 3580 80 593 EDETOX36 Phenethyl alcohol 60-12-8 12217 136a 22200a minus1765 172E-02 160E-02 35514 80 76 RIFM Db37 Benzyl alcohol 100-51-6 10814 110a 42900a minus2220a 603E-03 588E-03 25241 80 1497 EDETOX38 3 and 4-(4-Hydroxy-4-

methylpentyl)-3-cyclohexene-1-carboxaldehyde

31906-04-4 21032 256 6100a minus1896 127E-02 119E-02 7241 80 364 RIFM Db

39 Phenol 108-95-2 9411 146a 82800a minus2090a 813E-03 789E-03 65321 80 471 EDETOX40 Ethyl alcohol 64-17-5 4607 minus031a 1000000a minus3100a 794E-04 793E-04 79268 80 2117 EDETOX41 Geranyl nitrile 5146-66-7 14924 324 3732 minus0654 222E-01 109E-01 4052 80 505 RIFM Db42 Trichloromethyl phenyl carbinyl

acetate90-17-5 26754 343 3937 minus1907 124E-02 115E-02 452 40 5 RIFM Db

Acetic acid 6005 minus017a 1000000a minus2194 640E-03 628E-03 628304222-trichloro-1-phenylethanol 2255 286 21234 minus1849 142E-02 131E-02 2781



Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source

43 Methyl atrarate 4707-47-5 1962 256 110399 minus1729 187E-02 170E-02 18725 80 20 RIFM Db24-dihydroxy-36-dimethylbenzoicacid

18218 219 262843 minus1843 143E-02 133E-02 35088

Methanol 3204 minus077a 1000000a minus3300a 501E-04 501E-04 5006444 Lactic acid 50-21-5 9008 minus072a 1000000a minus2965 108E-03 108E-03 107861 80 1012 RIFM Db45 2-Methyl-2-propanol 75-65-0 7412 035a 1000000a minus1965 108E-02 105E-02 1046450 80 2 RIFM Db46 Triethanolamine 102-71-6 14919 minus100a 1000000a minus3875 133E-04 133E-04 13313 80 69 RIFM Db47 Diethyl maleate 141-05-9 17218 130 14000a minus2398 400E-03 392E-03 5491 80 54 RIFM Db

Maleic acid 11607 046a 7000a minus2376 420E-03 413E-03 2891Ethyl alcohol 4607 minus031a 1000000a minus3100a 794E-04 793E-04 792684-ethoxy-4-oxobut-2-enoic acid 14413 066 2656398 minus2556 278E-03 275E-03 72967

48 2-Ethyl-1-hexanol 104-76-7 13023 269 880a minus0853 140E-01 869E-02 7646 80 52 RIFM Db49 1-Decanol 112-30-1 15829 457a 37a minus1100a 794E-02 574E-02 212 40 002 RIFM Db50 Octanoic acid 124-07-2 14422 305a 789a minus1600a 251E-02 225E-02 1776 80 177 RIFM Db51 Lauric acid 143-07-7 20032 460a 48a minus0228 591E-01 140E-01 067 40 0164 RIFM Db52 Methyl 2-nonynoate 111-80-8 16824 357 1512 minus0628 235E-01 108E-01 1637 80 5 RIFM Db53 Estragole 140-67-0 14821 308 178a minus0764 172E-01 953E-02 1696 80 17 RIFM Db54 Diethyl phthalate 84-66-2 22224 242a 1080a minus2142 722E-03 693E-03 748 40 47 RIFM Db

phthalic acid 16613 073a 6970a minus2762 173E-03 171E-03 1195Ethyl alcohol 4607 minus031a 1000000a minus3100a 794E-04 793E-04 792682-(ethoxycarbonyl)benzoic acid 19419 180 97943 minus2282 522E-03 508E-03 4978

55 16 Hexanediol diglycidyl ether 16096-31-4 23031 108 378924 minus3251 561E-04 559E-04 2119 80 378 EDETOX56 17a-Hydroxyprogesterone 68-96-2 33047 317a 4293 minus3220a 603E-04 600E-04 026 40 1476 EDETOX57 1-Methoxypropan-2-ol 107-98-2 9012 minus021 1000000a minus2576 266E-03 263E-03 263055 80 229 EDETOX58 24-Dichlorophenoxyacetic acid 94-75-7 22104 281a 677a minus1831 147E-02 136E-02 920 40 12 EDETOX59 2-Butoxyethanol 111-76-2 11818 083a 1000000a minus3670a 214E-04 214E-04 21361 80 2739 EDETOX60 2-Ethoxyethanol 110-80-5 9012 minus032a 1000000a minus3600a 251E-04 251E-04 25096 80 178 EDETOX61 2-Isopropoxyethanol 109-59-1 10415 005a 1000000a minus3357a 440E-04 439E-04 43878 80 155 EDETOX62 2-Methoxyethyl acetate 110-49-6 11813 002 1000000a minus2734 185E-03 183E-03 183124 80 537 EDETOX

Acetic acid 6005 minus017a 1000000a minus2194 640E-03 628E-03 628304Ethylene glycol methyl ether 761 minus077a 1000000a minus2838 145E-03 144E-03 144371

63 2-Naphthylamine 91-59-8 14319 228a 11427 minus1315 484E-02 396E-02 4523 80 541 EDETOX64 2-Nitro-4-Phenylenediamine 5307-14-2 15314 053a 250472 minus2761 173E-03 172E-03 4309 80 217 EDETOX65 2-phenylphenol 90-43-7 17021 309a 700a minus1800a 158E-02 147E-02 1028 80 2667 EDETOX66 4-Acetamidophenol 103-90-2 15117 046a 14000a minus2791 162E-03 161E-03 2250 80 37 EDETOX67 4-Amino-2-nitrophenol 119-34-6 15413 096a 178553 minus2446 358E-03 352E-03 6285 80 451 EDETOX68 4-Aminobenzoic acid 150-13-0 13714 083a 6110a minus2344 453E-03 444E-03 2710 80 2837 EDETOX69 4-Aminophenol 123-30-8 10913 004a 16000a minus2613 244E-03 241E-03 3859 80 81 EDETOX70 4-Cyanophenol 767-00-0 11912 160a 175432 minus1547 284E-02 253E-02 44467 80 46 EDETOX71 4-Dimethylaminobenzene 60-11-7 2253 458a 369 minus0538 290E-01 108E-01 400 40 2157 EDETOX72 4-Heptyloxyphenol 13037-86-0 2083 439 1189 minus0483 329E-01 116E-01 1385 80 36 EDETOX73 4-Iodophenol 540-38-5 22001 291a 42717 minus1743 181E-02 164E-02 6992 80 28 EDETOX74 4-Nitroaniline 100-01-6 13813 139a 128579 minus1931 117E-02 111E-02 14316 80 48 EDETOX75 4-Nitrophenol 100-02-7 13911 191a 11600a minus1548 283E-02 251E-02 29117 80 41 EDETOX76 4-Pentyloxyphenol 18979-53-8 18025 350a 6936 minus0826 149E-01 842E-02 5844 80 29 EDETOX77 Acetylcysteine 616-91-1 16319 minus046 8885078 minus3628 235E-04 235E-04 20890 80 405 EDETOX78 Acetylsalicylic acid 50-78-2 18016 119a 4600a minus2140a 724E-03 698E-03 3212 80 312 EDETOX

Acetic acid 6005 minus017a 1000000a minus2194 640E-03 628E-03 6283042-Hydroxybenzoic acid 13812 226a 2240a minus1860a 138E-02 130E-02 2910

79 Androstenedione 63-05-8 28642 275a 3228 minus2649 225E-03 221E-03 071 40 1347 EDETOX80 Aniline 62-53-3 9313 090a 36000a minus1772 169E-02 159E-02 57284 80 376 EDETOX81 Atrazine 1912-24-9 21569 261a 347a minus1920 120E-02 113E-02 039 40 35 EDETOX82 Azodrin 6923-22-4 22317 minus020a 1000000a minus4141 722E-05 722E-05 7221 80 147 EDETOX83 Benzocaine 94-09-7 16519 186a 1310a minus1894 128E-02 120E-02 1575 80 48 EDETOX

4-Aminobenzoic acid 13714 083a 6110a minus2344 453E-03 444E-03 2710Ethyl alcohol 4607 minus031a 1000000a minus3100a 794E-04 793E-04 79268

84 Beta-estradiol 50-28-2 27239 401a 3087 minus2400a 398E-03 388E-03 120 40 38 EDETOX85 Boric Acid 10043-35-3 6183 017a 50000a minus3301a 500E-04 499E-04 2496 80 175 EDETOX86 Butachlor 23184-66-9 31186 450a 20a minus1620 240E-02 206E-02 041 40 44 EDETOX87 Caffeine 58-08-2 19419 minus007a 21600a minus3580a 263E-04 263E-04 567 40 326 EDETOX88 Carbaryl 63-25-2 20123 236a 110a minus1939 115E-02 108E-02 119 40 7391 EDETOX

N-methylcarbamate 7507 minus052 1412653 minus2638 230E-03 228E-03 3224581-Naphthol 14417 285a 866a minus0894 128E-01 803E-02 6954

89 Catechol 120-80-9 11011 088a 461000a minus1987 103E-02 988E-03 455662 80 18 EDETOX90 Chloramphenicol 56-75-7 32313 114a 2500a minus4304 497E-05 497E-05 012 40 29 EDETOX91 Cinnamyl anthranilate 87-29-6 2533 373 44 minus1512 308E-02 259E-02 114 40 533 EDETOX

Benzoic acid 2-amino- 13714 121a 3500a minus2056 879E-03 846E-03 2960Cinnamyl alcohol 13418 195a 71139 minus1459 347E-02 301E-02 21397

92 Deoxycorticosterone 64-85-7 33047 288a 3759 minus3350a 447E-04 445E-04 017 40 1255 EDETOX93 DFP 55-91-4 18415 117a 15400a minus2641 229E-03 226E-03 3479 80 243 EDETOX94 DHEA 53-43-0 28843 323a 3443 minus2308 492E-03 477E-03 164 40 1845 EDETOX95 Diazinon 333-41-5 30435 381a 1051 minus2056 880E-03 831E-03 087 40 141 EDETOX



deviation (RMSE) of most water solubility models are between 07and 10 log units The reason being that the average uncertainty inmeasuring water solubility values is no better than 06 log units(Hewitt et al 2009 Wang and Hou 2011) Considering the pre-diction accuracy of current QSAR models for water solubility thestandard deviation of the four different predictions is added to the

average value to give the consensus value used in the calculationsfor conservative purposes The final output then converts to thesaturated water solubility using Eq (7) below

C g cm S mol L MW g molwatersat μ 3 1000( ) = ( ) times ( ) times (7)

Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source

96 Diethylene glycol monobutyl etheracetate

124-17-4 20427 125 31000a minus2819 152E-03 150E-03 4660 80 4181 EDETOX

Acetic acid 6005 minus017a 1000000a minus2194 640E-03 628E-03 628304Diethylene glycol monobutyl ether 16223 056a 1000000a minus2845 143E-03 142E-03 141807

97 Dimethoate 60-51-5 22925 078a 23300a minus3469 340E-04 339E-04 790 40 26 EDETOX98 Dimethylnitrosamine 62-75-9 7408 minus057a 1000000a minus2663 217E-03 216E-03 215830 80 398 EDETOX99 Dinitrochlorobenzene 97-00-7 20255 217a 7228 minus2099 796E-03 763E-03 551 40 275 EDETOX100 Dipropylene glycol methyl ether 34590-94-8 1482 015 5717552 minus2994 101E-03 101E-03 57755 80 032 EDETOX101 Flutamide 13311-84-7 27622 335a 3153 minus2073 846E-03 802E-03 253 40 161 EDETOX102 Hippuric acid 495-69-2 17918 031a 3750a minus3235 582E-04 580E-04 218 40 12 EDETOX103 Lindane 58-89-9 29083 414a 337 minus1646 226E-02 197E-02 066 40 257 EDETOX104 Malathion 121-75-5 33035 236a 143a minus3463 344E-04 344E-04 005 10 448 EDETOX105 MbOCA 101-14-4 26716 391a 139a minus1541 288E-02 244E-02 034 40 59 EDETOX106 MDA 101-77-9 19827 159a 1000a minus2644a 227E-03 224E-03 224 40 329 EDETOX107 Methiocarb 2032-65-7 22531 292a 27a minus1798 159E-02 146E-02 039 40 154 EDETOX

N-methylcarbamate 7507 minus052 1412653 minus2638 230E-03 228E-03 3224584-(methylthio)-35-xylenol 16826 303 14688 minus1043 905E-02 623E-02 9157

108 Methyl-Parathion 298-00-0 26321 286a 377a minus2291 512E-03 496E-03 019 40 899 EDETOX109 NN-Diethyl-m-toluamide 134-62-3 19128 218a 10452 minus2889a 129E-03 128E-03 134 40 277 EDETOX110 Nicotinamide 98-92-0 12213 minus037a 500000a minus3078 836E-04 833E-04 41636 80 288 EDETOX111 Nicotinic acid 59-67-6 12311 036a 18000a minus4620a 240E-05 240E-05 043 40 33 EDETOX112 Nitrobenzene 98-95-3 12311 185a 2090a minus1405 394E-02 337E-02 7049 80 411 EDETOX113 N-Phenyl-2-naphthylamine 135-88-6 21929 438a 171 minus0619 240E-01 101E-01 173 40 0 EDETOX114 n-Propoxyethanol 2807-30-9 10415 028 2766637 minus2370 427E-03 420E-03 116097 80 33 EDETOX115 o-cresyl glycidyl ether 2210-79-9 16421 210 20957 minus1701 199E-02 181E-02 3798 80 102 EDETOX116 o-toluidine 95-53-4 10716 132a 16600a minus1619 241E-02 220E-02 36455 80 498 EDETOX117 Paraoxon 311-45-5 2752 198a 3640a minus3101 793E-04 789E-04 287 40 1552 EDETOX118 Pentachlorophenol 87-86-5 26634 512a 14a minus0613 244E-01 964E-02 135 40 15 EDETOX119 Phosmet 732-11-6 31732 278a 244a minus2990 102E-03 102E-03 002 10 215 EDETOX120 Phoxim 14816-18-3 2983 439a 41a minus1544 286E-02 240E-02 010 10 29 EDETOX121 Pirimicarb 23103-98-2 23829 170a 2700a minus2878 133E-03 132E-03 355 40 282 EDETOX

Carbamic acid dimethyl- 8909 minus029 1248465 minus2626 237E-03 235E-03 2929942-(dimethylamino)-56-dimethylpyrimidin-4-ol

16721 163 479893 minus2091 811E-03 779E-03 37393

122 Progesterone 57-83-0 31447 387a 88a minus2820a 151E-03 150E-03 001 10 071 EDETOX123 Propoxur 114-26-1 20925 152a 1860a minus2590a 257E-03 253E-03 471 40 196 EDETOX

N-methylcarbamate 7507 minus052 1412653 minus2638 230E-03 228E-03 3224582-isopropoxyphenol 15219 209a 47046 minus1566 272E-02 241E-02 11331

124 Propylene glycol 57-55-6 761 minus092a 1000000a minus2952 112E-03 111E-03 111203 80 138 EDETOX125 Testosterone 58-22-0 28843 332a 2701 minus2660a 219E-03 216E-03 058 40 348 EDETOX126 Theophylline 58-55-9 18017 minus004a 7360a minus3512 307E-04 307E-04 226 40 169 EDETOX127 Thiourea 62-56-6 7612 minus108a 142000a minus3074 843E-04 841E-04 11943 80 34 EDETOX128 Trichlorocarbanilide 101-20-2 31559 474 175 minus1485 328E-02 268E-02 047 40 0392 EDETOX129 Trichloromethane 67-66-3 11938 197a 7950a minus1796a 160E-02 150E-02 11916 80 82 EDETOX130 Triclopyr 55335-06-3 25647 276 440a minus2286 518E-03 502E-03 221 40 165 EDETOX131 Trimethylamine 75-50-3 5911 minus038a 1630000a minus3720a 191E-04 190E-04 31042 80 6 EDETOX

a Experimental determined data Kpcorr is the corrected Kp calculated using Eq (5) Jmax is calculated using Eq (6) The metabolites of esters ie carboxylic acid and alcohol

moieties are also listed below the parent material shown as italic

Table 3The seven in silico models employed to predict a consensus (average) Kow value when experimental data are lacking

Prediction model Moleculardescriptor

Modelingmethod

Training sets Developer Remark Web addressreference

n r2

AlogPSVCCLAB Atom-based ANN 12908 095 VCCL Free httpwwwvcclaborglabalogpsAlogPPipeline Pilot Atom-based MLR 8364 090 Accelrys Commercial httpaccelryscomproductspipeline-pilotVGlogPJCHEM Atom-based MLR 893 093 Chemaxon Commercial httpwwwchemaxoncomXLOGP3 Atom-based MLR 8199 091 SIOC Free httpwwwsioc-ccbgaccnp=42ampsoftware=xlogp3KOWWINEPI SUITE Fragment-based MLR 2447 098 EPA Free httpwwwepagovopptintrexposurepubsepisuitehtmKLogPJCHEM Fragment-based MLR 1663 093 Chemaxon Commercial httpwwwchemaxoncomLogPMultiCase Fragment-based MLR gt8000 094 Multicase Inc Commercial httpwwwmulticasecom

All these models calculate Kow based on molecular structure The training sets of these models cover MW ranging from 18 to 992 n = number of substances r2 = squarecorrelation coefficient


24 Determining the percent skin absorption based oncalculated Jmax

According to Kroes et al the skin absorption (A) of a materialcan be estimated based on its calculated Jmax as shown in Eq (8)

A JJ

JJ

max

max

max

max

( ) =le

lt legt

⎧⎨⎪

⎩⎪

10 0 140 0 1 1080 10

(8)

It has been shown that the esters such as salicylate readilyhydrolyse or metabolize to corresponding carboxylic acid and alcoholon the skin (Belsito et al 2007) Taking this into consideration forany ester the metabolites which are the carboxylic acid and alcoholas well as the parent ester will be processed through the work-flow individually in order to obtain their individual Jmax values Thefinal Jmax value to use in assigning a default skin absorption valuerepresentative of the ester material will be based on the chemicalentity that gave the highest value to err on the conservative sideThe overall workflow of conducting skin absorption prediction isdepicted in Fig 1

25 The conservative nature of the calculations

A hallmark of the method employed in this exercise is conser-vancy This conservancy is demonstrated in the calculations of log Kowlog S and log Kp where multiple models are used to determine a meanvalue andor standard error of estimation are incorporated into the cal-culations Conservancy is also invoked in the 80 40 and 10 categorycutoff values for percent skin absorption Specifically a review of theexperimental percent skin absorption values reported in Table 2 revealsthat none of the 131 validation materials have skin absorption gt80in fact only four materials (ie benzyl acetate benzyl benzoate car-baryl and cinnamyl alcohol) have skin absorptions gt 65 Additionally26 materials exhibit absorption between 65 and 35 57 materialsexhibit absorption between 35 and 10 while 44 materials revealabsorptions of lt10

3 Model validation

31 Validation and refinement of Potts and Guyrsquos skin permeabilitycoefficient Kp model

The original equation to predict skin permeability coefficient wasdeveloped by Potts and Guy (1992) This model was developed onFlynnrsquos data set with 93 materials which were not entirely com-posed of fragrance or ldquofragrance-likerdquo material Therefore wequestioned the suitability of this model and resolved to tailor it tofit fragrance and ldquofragrance-likerdquo materials

One hundred five (105) fragrance and ldquofragrance-likerdquo materi-als (ie materials with MW log Kow and log S falling into thefragrance chemical space) compose data set I (Table 1) All of themhave experimentally determined Kp and log Kow The estimated

log Kp are calculated using Potts and Guyrsquos equation (Eq (2)) Thesquare of linear correlation coefficient (r2) between experimentaland estimated log Kp is 0578 with a standard error (SE) of 0712(Fig 3A) Such results indicated that Potts and Guyrsquos equation wasacceptable for fragrance materials However Potts and Guyrsquos equa-tion was developed for chemicals of diverse structures andfunctionality in general Therefore 105 fragrance and ldquofragrance-likerdquo materials in data set I were used to re-fit the parameters ofPotts and Guyrsquos equation

Table 4The four in silico models employed to predict a consensus (average) water solubility (log S) value when experimental data are lacking


Modelingmethod


n r2

WSKOWEPI SUITE Physicochemicalproperty

MLR 1450 097 EPA Free httpwwwepagovopptintrexposurepubsepisuitehtm

SolubilityPipeline Pilot E-state indices ANN 1291 091 Accelrys Commercial httpaccelryscomproductspipeline-pilotKLogSMultiCase Fragment-based MLR 483 095 Multicase Inc Commercial httpwwwmulticasecomXLOGS Fragment-based Read-across 4171 082 SIOC Free httpwwwsioc-ccbgaccnp=42ampsoftware=xlogs

All these models calculate log S based on molecular structure The training sets of these models cover MW ranging from 27 to 667 and log Kow minus85 to 10 n = number ofsubstances r2 = square correlation coefficient

Fig 3 The correlation of experimental log Kp versus estimated log Kp (A) Log Kp

calculated based on Potts and Guyrsquos equation (Eq (2)) (B) Log Kp calculated basedon Eq (3)


The updated Eq (3) namely RIFMrsquos equation has the r2 of 0703and SE of 0594 on data set I which is a fragrance focused data set(Fig 3B) The higher r2 value indicates a better fit of Potts and Guyrsquosequation for the fragrance chemical domain Therefore RIFMrsquos equa-tion is used to calculate Kp for fragrance and ldquofragrance-likerdquomaterials

As mentioned above Kp is not the final output of the SAM Ratherit is an intermediate parameter used to calculate the Jmax (see Eq(1)) Since the conservative estimation is always favorable we addthe standard error of 0594 of log Kp calculated from RIFMrsquos equa-tion as shown in Eq (4)

32 Validation of the SAM

Validation is crucial to prove the suitability and accuracy of amodel The validation data set (data set II) contains 131 fragranceand ldquofragrance-likerdquo materials ie their MW log Kow and log S fellwithin the fragrance defined ranges discussed above with knownexperimental skin absorption data (Table 2) The Jmax was

obtained following the workflow described above and as depictedin Fig 1 The calculated Jmax values were plotted in relation to theexperimental absorption values for these ingredients as shown inFig 4A The solid blue dots represent the fragrance materials andthe blue circles represent the ldquofragrance-likerdquo materials Of these131 materials 21 were esters (solid red and open red circles) TheJmax of the esters was obtained as described in section 24 By com-paring the actual experimentally derived absorption data with theassigned default values based on Jmax one can clearly see that allof the materials fall below the Kroes et al (2007) predicted defaultabsorption value boundaries This means the SAM works very wellfor the materials in our validation set It is worth noting that all theesters happen to be under the 80 area This is because we take themetabolites of esters (ie carboxylic acid and alcohol moieties) intoconsideration and choose the most conservative Jmax to estimate theskin absorption percentage As mentioned in section 22 the vali-dation data set contains data from in vitro and in vivo data fromhumans monkeys and pigs None of the materials have the exper-imental determined skin absorption percentage exceeding the

Fig 4 Plots of calculated Jmax versus experimental percent skin absorption Shaded areas represent absorption zones (10 (light blue) 40 (yellow) and 80 (green)) fromEquation (8) (A) The Jmax of 131 fragrance (blue and red dots) and ldquofragrance-likerdquo materials (blue and red circles) were calculated through the workflow depicted in Fig 1(B) Assuming there are no experimentally derived parameters available use consensus values to estimate log Kow and log S and using Eq (4) to calculate the Kp most ofthe calculated Jmax were shifted right


predictions by SAM This fits the objective of SAM ie giving con-servative estimations for human in vivo skin absorption instead ofusing the default value of 100

As shown in Fig 1 when experimentally determined param-eters ie log Kow water solubility or Kp are available those parametersshould be used Our validation data set contained 96 materials withexperimentally determined log Kow 82 with experimentally deter-mined water solubility (Cwater

Sat) and 27 had an experimentallydetermined Kp Twenty-two (22) had all three parameters experi-mentally derived (Table 2 marked with a) To further test the SAMand understand how skin absorption will map in a case wherethere are no experimental data available for any of the param-eters to calculate Jmax we used a predicted log Kow Cwater

Sat andlog Kp to calculate the Jmax of our validation set When we plottedthe calculated Jmax results most of the materials shifted to the right(Fig 4B) That is using all predicted parameter values in onersquos Jmax

calculation usually results in a more conservative dermal absorp-tion prediction ndash preferred when experimental data are lacking

It should be noted that due to the inherent conservatism builtinto the SAM by adding the deviation of 0594 to the calculated logKp and the standard deviation to the average log S there is no needto be extra conservative when dealing with range limit Jmax values

33 Meeting the OECD principles for QSAR validation

All the QSAR models used in this exercise meets the Organisationfor Economic Co-operation and Development (OECD) principles forvalidation for regulatory purposes (OECD 2007) Specifically themodels having 1) a defined endpoint 2) an unambiguous algo-rithm 3) a defined domain of applicability 4) appropriate measuresof goodness-of-fit robustness and predictivity and 5) a mechanis-tic interpretation

34 Applicability domain of the SAM

Our proposed fragrance SAM includes three QSAR models ndash spe-cifically log Kow log S and log Kp It is therefore necessary to ensurethat onersquos target material falls within the application domain whenemploying each model For log Kow the training sets coveredmaterials with molecular weight between 18 and 992 (Table 3)Of these log Kow models four models were trained by a data setcontaining more than 8000 materials For example the trainingset of AlogPS contains almost 13000 structurally diverse materi-als and covers the fragrance chemical space The final log Kow is theaverage value of these seven models For the 105 fragrance andldquofragrance-likerdquo materials in the data set I the squared correla-tion coefficient (r2) of the experimental and estimated log Kow are097 and the root-mean-square deviation (RMSE) is 025 This meansthat the consensus log Kow model works very well with fragrancematerials

The training sets for log S models covered materials with molec-ular weight from 27 to 667 and log Kow from minus85 to 10 (Table 2) Thefragrance chemical space is also covered by these training sets Howeverthe number of materials in each training set is comparably less thanthose of log Kow models The largest training set is that of XLOGS whichcontains 4171 compounds with known log S values The r2 and RMSEin regression were 082 and 096 log units respectively (Duan et al2012) As discussed above the uncertainty of the water solubility is sig-nificant in experiments The prediction models usually have high RMSETherefore the estimated log S from our model was added by the stan-dard deviations from different prediction models for conservativepurpose

Kp was calculated based on RIFMrsquos equation (Eq (3))which is an updated formula from Potts and Guyrsquos equation(Eq (2)) The training set (data set I) contains 105 fragrance andldquofragrance-likerdquo materials which makes it a fragrance focused model

The log Kp model covered materials with MW from 33 to 330 andlog Kow from minus1 to 6

If the target material falls into these training set ranges it is con-sidered to be in the fragrance application domain and thereforesuitable for processing as we have described If the target materi-al does not fall into this application domain it may still be processedbut a low confidence caution should be given (ie Low Confi-dence Warning in Fig 1)

4 Application of the SAM in safety assessment

We have demonstrated that the SAM is sound to use when ex-perimental skin absorption data are lacking and thus we proposeusing the outcomes of this semi-quantitative mechanistic insilico model instead of the unreasonable 100 default in caseswhere it will make a difference in thresholds like TTC (Thresholdof Toxicological Concern) and MOE or MOS (margin of exposureor margin of safety) (Eaton and Gilbert 2007 Kroes et al 20052007)

To demonstrate this consider the following hypothetical exampleMaterial ldquoXrdquo that meets the above described criteria (see section21) does not have animal data or suitable analogs for read-acrossto clear it for the toxicity endpoint of interest The material is usedat very low levels and its TTC of 30 μgkgday (based on a CramerClass I material) is considered The total systemic exposure basedon the default assumption of 100 skin absorption is 34 μgkgday These values are so close that a more representative percentabsorption may bring the total systemic exposure below the TTCleading to different risk management The following steps are takento calculate absorption based on Jmax

1 Get the MW (gmol) and check for experimentally determinedparameters of Kp (cmh) Cwater

sat (mgL) and log Kow In this casewe know the structure and the MW = 204 gmol but have no otherexperimental values

2 Determine log Kow by calculating it from the seven in silico toolsof Table 1

log Kow model Result

KOWWINEPI SUITE 548AlogPPipeline Pilot 418VGlogPJCHEM 398KLogPJCHEM 427LogPMultiCase 422XLOGP3 508AlogPSVCCLAB 465Average plusmn STD 455 plusmn 055

3 Calculate Kp (cmh) by using Eq (4) That is log Kp (cmh) = minus1356 + 0759 times log Kow minus 00118 times MW = minus1356 + (0759 times455) minus (00118 times 20436) = minus0314 Note that the deviation of 0594has been added to the calculated log Kp for conservative pur-poses Therefore Kp = 10log Kp = 10minus0314 = 0485 cmh

4 Correct the Kp by using Cleek and Bundgersquos equation (Eq (5))

Kpcorr KpKp MW= +

⎛⎝⎜

⎞⎠⎟

= + times⎛⎝⎜

⎞⎠⎟

=

12 6

0 485 10 485 204 36

2 60 13

22 cm h5 Derive water solubility (Cwater

Sat) by determining it from the fourin silico tools of Table 3

log S model Result

WSKOWEPI SUITE minus591SolubilityPipeline Pilot minus448KLogSMultiCase minus430XLOGS minus425Average plusmn STD minus473 plusmn 079


This water solubility is used to calculate Jmax For conservativepurposes the standard deviation is added to the average log Sminus473 + 079 = minus394 S = 10 times log S = 10 minus 394 = 0000114 molL S is converted to Cwater

Sat = S times MW times 1000 = 0000114 times 20436 times1000 = 2329 μgcm3

6 Calculate Jmax as follows Jmax = Kpcorr times Cwat

sat = 0132 times 2329 =307 μgcm2h According to Eq (8) the 40 estimated percentskin absorption is assigned to this material By assigning 40 ab-sorption the systemic exposure becomes 1522 μgkgday whichis approximately twofold below its determined TTC and nowclears the toxicological endpoint of interest

This example demonstrates the unreasonableness of default-ing to 100 skin absorption when experimental data are lacking andthe usefulness of this model in calculating more representativepercent absorption values for fragrance or ldquofragrance-likerdquomaterials

It is worth noting that even though this example uses seven dif-ferent log Kow prediction models and four different water solubilitymodels to obtain the consensus estimated log Kow and water sol-ubility values one does not have to use as many ndash one may evenuse other validated models to get a consensus estimation Forexample if only the three free accessible log Kow models(ie KOWWIN XlOGP3 and ALOGPS) and two free accessiblewater solubility models (WSKOW and XLOGS) are used the finalconclusion would not change (Jmax = 400 μgcm2h in 40 absorp-tion range)

5 Conclusion

We concur with Kroes et alrsquos (2007) assumption that 100 skinabsorption is not a scientifically supportable default value in theabsence of experimental data Furthermore we showed that Kroeset alrsquos three tier percent skin absorption scheme for cosmeticingredients fit very well for fragrance ingredients We developeda specific skin absorption model namely SAM for fragranceingredients whose default absorption value is either 10 40or 80 To which percent absorption domain a fragrance ingredi-ent falls depends on its calculated Jmax Our in silico model wasvalidated using 131 fragrance materials with experimental in vitroor in vivo absorption data obtained from either human or pigskin Although the SAM deviates from the 100 default absorp-tion value it still has a lot of conservatism built into it A schematicof the SAM is shown in Fig 1 and an example of how to use themodel is discussed in Section 4 This model is very useful in pri-oritizing testing when the estimated exposure is very close to TTCor when it will make a significant difference in estimating the pointof departure

Funding source

This study was part of an internal research program of the Re-search Institute for Fragrance Materials Inc No third party fundingwas involved

Conflict of interest

The authors declare that there are no conflicts of interest

Transparency document

The Transparency document associated with this article can befound in the online version

Acknowledgment

The authors would like to thank Dr David K Wilcox and Dr AnneMarie Api for their support and guidance

References

Aggarwal M Battalora M Fisher P Huumlser A Parr-Dobrzanski R Soufi M et al2014 Assessment of in vitro human dermal absorption studies on pesticides todetermine default values opportunities for read-across and influence of dilutionon absorption Regul Toxicol Pharmacol 68 412ndash423

Api AM Ritacco G Hawkins DR 2013 The fate of dermally applied [14C]d-limonene in rats and humans Int J Toxicol 32 130ndash135

Barber ED Teetsel NM Kolberg KF Guest D 1992 A comparative study of therates of in vitro percutaneous absorption of eight chemicals using rat and humanskin Fundam Appl Toxicol 19 493ndash497

Belsito D Bickers D Bruze M Calow P Greim H Hanifin JM et al 2007 Atoxicologic and dermatologic assessment of salicylates when used as fragranceingredients Food Chem Toxicol 45 S318ndashS361

Belsito D Bickers D Bruze M Calow P Dagli ML Fryer AD et al 2011 Atoxicological and dermatological assessment of macrocyclic lactone and lactidederivatives when used as fragrance ingredients Food Chem Toxicol 49 (Suppl2) S219ndashS241

Belsito D Bickers D Bruze M Calow P Dagli ML Dekant W et al 2012a Atoxicologic and dermatologic assessment of cyclopentanones andcyclopentenones when used as fragrance ingredients Food Chem Toxicol 50(Suppl 3) S517ndashS556

Belsito D Bickers D Bruze M Calow P Dagli ML Fryer AD et al 2012b Atoxicological and dermatological assessment of aryl alkyl alcohol simple acid esterderivatives when used as fragrance ingredients Food Chem Toxicol 50 (Suppl2) S269ndashS313

Belsito D Bickers D Bruze M Calow P Dagli ML Fryer AD et al 2012c Atoxicological and dermatological assessment of aryl alkyl alcohols when usedas fragrance ingredients Food Chem Toxicol 50 (Suppl 2) S52ndashS99

Bickers D Calow P Greim H Hanifin JM Rogers AE Saurat JH et al 2005A toxicologic and dermatologic assessment of cinnamyl alcohol cinnamaldehydeand cinnamic acid when used as fragrance ingredients Food Chem Toxicol 43799ndash836

Bobin MF Durand F Pelletier J Martini MC 1997 Dosage of farnesol in the skinIn Agropolis VE (Ed) International Symposium on Flavours and Sensory RelatedAspects Cernobbio Villa Erba pp 244ndash250

Bronaugh RL Wester RC Bucks D Maibach HI Sarason R 1990 In vivopercutaneous absorption of fragrance ingredients in rhesus monkeys and humansFood Chem Toxicol 28 369ndash373

Cleek R Bunge A 1993 A new method for estimating dermal absorption fromchemical exposure 1 General approach Pharm Res 10 497ndash506

Cross SE Anderson C Roberts MS 1998 Topical penetration of commercialsalicylate esters and salts using human isolated skin and clinical microdialysisstudies Br J Clin Pharmacol 46 29ndash35

Duan BG Li Y Li J Cheng TJ Wang RX 2012 An empirical additive model foraqueous solubility computation success and limitations Acta Phys Chim Sin28 2249ndash2257

Eaton DL Gilbert SG 2007 Principles of toxicity In Klaassen CD (Ed) Casarettamp Doullrsquos Toxicology The Basic Science of Poisons seventh ed McGraw-HillEducation Boston pp 11ndash44

EFSA Panel on Plant Protection Products and their Residues 2011 Scientific opinionon the science behind the revision of the guidance document on dermalabsorption EFSA J 9 2294

EFSA Panel on Plant Protection Products and their Residues 2012 Guidance onDermal Absorption EFSA J 10 2665

European Commission 2004 Guidance document on dermal absorption (ltSanco2222000 rev 7 19 March 2004 (httpeceuropaeufoodplantprotectionevaluationguidancewrkdoc20_rev_enpdfgt) accessed Feb 3 2014)

Fick A 1855 V On liquid diffusion Philos Mag 10 30ndash39 ISO 4Flynn GL 1990 Physicochemical determinants of skin absorption In Gerrity T

R Henry C J (Eds) Principles of Route-to-route Extrapolation for RiskAssessment Proceedings of the Workshops on Principles of Route-to-RouteExtrapolation for Risk Assessment Held March 19ndash21 1990 in Hilton Head SouthCarolina and July 10ndash11 1990 in Durham North Carolina Elsevier New Yorkpp 93ndash127

Ford RA Hawkins DR Schwarzenbach R Api AM 1999 The systemic exposureto the polycyclic musks AHTN and HHCB under conditions of use as fragranceingredients evidence of lack of complete absorption from a skin reservoir ToxicolLett 111 133ndash142

Ford RA Hawkins DR Mayo BC Api AM 2001 The in vivo dermal absorptionand metabolism of [4-14C] coumarin by rats and by human volunteers undersimulated conditions of use in fragrances Food Chem Toxicol 39 153ndash162

Gilpin S Hui X Maibach H 2010 In vitro human skin penetration of geraniol andcitronellol Dermatitis 21 41ndash48

Green D Jones C Williams S 2014 In Vitro Human Skin Penetration of MethylN-Methylanthranilate (Dimethyl Anthranilate) RIFM Report 66570

Green DM Brain K 2005 In Vitro Human Skin Penetration of Fragrance MaterialGeranyl Nitrile RIFM Report 47379


Green DM Brain SI 2001 In-Vitro Human Skin Penetration of Methyl EugenolEstragole Acetyl Cedrene and OTNE RIFM Report 37084

Green DM Walters KA 2006 In Vitro Human Skin Penetration of FragranceMaterial HMPCC (3 and 4-(4-hydroxy-4-methylpentyl)-3-cyclohexene-1-carboxaldehyde) under Both In-use and Occluded Conditions RIFM Report54458

Green DM Jones C Walters KA 2008 In Vitro Human Skin Penetration ofFragrance Material Dihydromyrcenol under Occluded Conditions from a 70Ethanol Vehicle RIFM Report 55341

Guy RH 2010 Predicting the rate and extent of fragrance chemical absorption intoand through the skin Chem Res Toxicol 23 864ndash870

Hawkins DR Kirkpatrick D Aikens PJ Whittle EB Saxton JE 1995 DermalAbsorption of (14)C-trichloromethyl Phenyl Carbinyl Acetate in Man RIFM Report23914

Hawkins DR Elsom LF Kirkpatrick D Ford RA Api AM 2002 Dermal absorptionand disposition of musk ambrette musk ketone and musk xylene in humansubjects Toxicol Lett 131 147ndash151

Hawkins DW Forrest TJ Moore DH McTigue JJ 1994 The Dermal Absorptionof (14)C-BMHCA (14)C-para-tert-butyl-alpha-Methylhydrocinnamaldehyde inMan RIFM Report 23500

Hewitt M Cronin MTD Enoch SJ Madden JC Roberts DW Dearden JC 2009In silico prediction of aqueous solubility the solubility challenge J Chem InfModel 49 2572ndash2587

Hostynek JJ 2008 Permeability of human skin to metal compounds with focuson nickel and copper In Zhai H Wilhelm KP Maibach HI (Eds) Marzulliand Maibachrsquos Dermatotoxicology seventh ed Taylor amp Francis New York pp215ndash226

Hotchkiss SAM 1998 Absorption of fragrance ingredients using in vitro modelswith human skin In Frosch P Johansen J White I (Eds) Fragrances SpringerBerlin Heidelberg pp 125ndash135

Isola DA Api AM 2002 In vitro human skin penetration of seven radiolabelledfragrance materials Toxicologist 66 165

Jimbo Y 1983 Penetration of fragrance compounds through human epidermis JDermatol 10 229ndash239

Judson RS Martin MT Egeghy P Gangwal S Reif DM Kothiya P et al 2012Aggregating data for computational toxicology applications the USEnvironmental Protection Agency (EPA) aggregated computational toxicologyresource (ACToR) system Int J Mol Sci 13 1805ndash1831

Keith DE 2003 Bathing the term newborn personal cleanser considerations InHoath SB Maibach HI (Eds) Neonatal Skin Structure and Function seconded Taylor amp Francis New York pp 211ndash238

Kraeling MEK Bronaugh RL 1997 In vitro percutaneous absorption of alphahydroxy acids in human skin J Soc Cosmet Chem 48 187ndash197

Kraeling MEK Bronaugh RL 2003 In vitro absorption of triethanolamine throughhuman skin Cutan Ocular Toxicol 22 137ndash145

Kroes R Kleiner J Renwick A 2005 The threshold of toxicological concern conceptin risk assessment Toxicol Sci 86 226ndash230

Kroes R Renwick AG Feron V Galli CL Gibney M Greim H et al 2007Application of the threshold of toxicological concern (TTC) to the safety evaluationof cosmetic ingredients Food Chem Toxicol 45 2533ndash2562

Liu FC Hotchkiss SAM 1997 Comparative Skin Absorption Metabolism andProtein Binding of the Sensitizers Eugenol and Isoeugenol RIFM Report 30499

Madsen JT Vogel S Johansen JD Sorensen JA Andersen KE Nielsen JB 2011Percutaneous penetration characteristics and release kinetics of contact allergensencapsulated in ethosomes Cutan Ocular Toxicol 30 38ndash44

Magnusson BM Anissimov YG Cross SE Roberts MS 2004 Molecular size asthe main determinant of solute maximum flux across the skin J InvestigDermatol 122 993ndash999

Mannhold R Poda GI Ostermann C Tetko IV 2009 Calculation of molecularlipophilicity state-of-the-art and comparison of log P methods on more than96000 compounds J Pharm Sci 98 861ndash893

Ngo MA OrsquoMalley M Maibach HI 2010 Percutaneous absorption and exposureassessment of pesticides J Appl Toxicol 30 91ndash114

Politano VT Diener RM Christian MS Hawkins DR Ritacco G Api AM 2013The pharmacokinetics of phenylethyl alcohol (PEA) safety evaluationcomparisons in rats rabbits and humans Int J Toxicol 32 39ndash47

Potts R Guy R 1992 Predicting skin permeability Pharm Res 9 663ndash669Smith CK Moore CA Elahi EN Smart AT Hotchkiss SA 2000 Human skin

absorption and metabolism of the contact allergens cinnamic aldehyde andcinnamic alcohol Toxicol Appl Pharmacol 168 189ndash199

Sundberg JP Nanney LB Fleckman P King LE 2012 23 ndash skin and adnexa InTreuting PM Dintzis SM (Eds) Comparative Anatomy and Histology AcademicPress San Diego pp 433ndash455

Tetko IV Poda GI Ostermann C Mannhold R 2009 Accurate in silico log Ppredictions one canrsquot embrace the unembraceable QSAR Comb Sci 28 845ndash849

Tonge RP 1995 The Cutaneous Disposition of the Sensitizing ChemicalsHydroxycitronellal and Dinitrochlorobenzene RIFM Report 556

van Ravenzwaay B Leibold E 2004 The significance of in vitro rat skin absorptionstudies to human risk assessment Toxicol In Vitro 18 219ndash225

Wang J Hou T 2011 Recent advances on aqueous solubility prediction CombChem High Throughput Screen 14 328ndash338

Watkinson AC Brain KR Walters KA Hadgraft J 1992 Prediction of thepercutaneous penetration of ultra-violet filters used in sunscreen formulationsInt J Cosmet Sci 14 265ndash275

Williams F 2004 EDETOX Evaluations and predictions of dermal absorption of toxicchemicals Int Arch Occup Environ Health 77 150ndash151

Yang JJ Krueger AJ Roy TA Neil W Mackerer CR 1994 In vitro percutaneousabsorption of gasoline oxygenates in the rat and human Toxicologist 14 104

Yano T Nakagawa A Tsuji M Noda K 1986 Skin permeability of variousnon-steroidal anti-inflammatory drugs in man Life Sci 39 1043ndash1050


An in silico skin absorption model for fragrance materials

Introduction

Methodology development and data sets

Defining the chemical space of fragrance materials

Datasets for validating and refining SAM

Calculating Jmax

Calculate Kp

Calculate log Kow

Calculate water solubility (CwaterSat)

Determining the percent skin absorption based on calculated Jmax

The conservative nature of the calculations

Model validation

Validation and refinement of Potts and Guys skin permeability coefficient Kp model

Validation of the SAM

Meeting the OECD principles for QSAR validation

Applicability domain of the SAM

Application of the SAM in safety assessment

Conclusion

Funding source



Acknowledgment

References

is maintained in a saturated solution or at steady state equilibri-um whose flux describes the amount of permeant per unit time andarea (ie μgcm2h) (Magnusson et al 2004) Jmax is independentof the formulation in which the material contacts the skin and is aconstant value when the formulation does not change the skin barrier(Kroes et al 2007) Based on Jmax the default absorption valuesproposed were as follows

bull Material with Jmax le01 μgcm2h should be assigned a skin ab-sorption default value of less than 10

bull If the Jmax value is gt01 μgcm2h but le10 μgcm2h the defaultskin absorption assigned did not exceed 40

bull If a material had a Jmax of gt10 μgcm2h the default skin absorp-tion assigned was no more than 80

The three default skin absorption percentages were proposedto represent low absorbed medium absorbed and high absorbedmaterial These default values were derived from a broad rangeof 15 cosmetic ingredients and considered several worst-case as-sumptions such as (i) cosmetic ingredients are present at saturationlevels (ii) no depletion of the ingredient occurs during the expo-sure period (iii) the formulation does not affect the skin barrierand (iv) by using the maximal flux over the entire exposure timethe lower flux during the lag time is ignored (Kroes et al 2007)In another study Guy proved that the skin absorption of chemi-cals may be classified by their calculated Jmax (Guy 2010) Usingthe same model the Jmax of 20 fragrance materials were calculatedand 16 of them were taken one step further to calculate their ab-sorption percentage by including the applied dose area and timeOver-prediction was observed in 14 materials and thereforethis approach could also be considered an extreme estimation ofabsorption

Based on the abovementioned studies it is apparent that the keyto determining a substancersquos default absorption value is to get anaccurate Jmax Centered on this reasoning we propose an in silico semi-quantitative mechanistic model for assigning the same default skinabsorption values as proposed by Kroes et al (2007) but specifi-

cally constructed around the fragrance material (Fig 1) As we willshow their overall methodology can be applied to derive similarabsorption values for RIFMrsquos fragrance materials that lack experi-mental skin absorption data provided we tailor the model tospecifically fit a defined set of fragrance material physicochemicalparameters Our mechanistic model was validated with a fragrance-focused data set containing 131 materials All resulted in predictedvalues fitting the proposed three-tiered SAM based on Jmax ranges

2 Methodology development and data sets

21 Defining the chemical space of fragrance materials

Getting insight into the chemical space of fragrance materialsand where this space is located in the chemical universe was in-structive for us to get the big picture and develop a fragrance-focused SAM Herein 2120 fragrance materials with known chemicalstructure in RIFM database were reviewed The fragrance chemi-cal space was profiled using three physicochemical properties thatsignificantly influence the overall absorption of a topically appliedsubstance namely molecular weight (MW) log Kow (octanolwater partition coefficient) and water solubility (S) For perspectivethese were compared to more than half a million industry chemi-cals from the United States Environmental Protection Agency (USEPA)ACTOR database (Judson et al 2012)

As shown in Fig 2 the MW log Kow and log S values of ~500000industry chemicals from USEPA ACTOR database are calculatedusing EPI Suite (USEPA) and plotted to represent the chemicaluniverse (Fig 2A blue dots) The same parameters of 2120fragrance materials are also calculated and plotted in the samechart (Fig 2A red dots) Clearly not only the number of fragrancematerial is significantly small but the fragrance space is limitedFurther analysis indicates that more than 99 of fragrance mate-rials gave a MW ranging from 30 to 330 (Fig 2B) a log Kow fromminus1 to 9 (Fig 2C) and log S from minus9 to 1 (Fig 2D) As such we con-sider any materials falling within these ranges as ldquofragrance-likerdquomaterials

Fig 1 Workflow for applying the skin absorption model (SAM) in safety assessment For a target material first look for available skin absorption data If an experimentalvalue is not available look for experimentally derived Kp and water solubility values to calculate Jmax If an experimentally derived Kp is not available look for an experi-mentally derived log Kow or use consensus value to estimate the log Kow to determine a predicted Kp using Eq (4) This predicted Kp may be used with experimentallyderived or predicted solubility values to calculate Jmax Then the percent skin absorption is estimated based on Eq (8) If the fragrance material is an ester oneneeds to calculate the Jmax of the parent and breakdown products (ie carboxylic acid and alcohol moieties) For conservative purposes accept the value that gives the highestskinabsorption






23 Calculating Jmax




231 Calculate Kp



(n = 93 r2 = 067 SE = 0794 F = 8405)















K cm hK

K MWpcorr p

p

( ) =+

times1

2 6

(5)









Table 1 (continued)








(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source



54464-57-2 23438 431 60 minus0846 143E-01 775E-02 465 40 1651 RIFM Db





1222-05-5 25841 590a 18a 0079 120E+00 143E-01 025 40 516 RIFM Db









1205-17-0 19222 214 1390 minus1996 101E-02 957E-03 1330 80 501 RIFM Db










31906-04-4 21032 256 6100a minus1896 127E-02 119E-02 7241 80 364 RIFM Db






Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source


18218 219 262843 minus1843 143E-02 133E-02 35088





















Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source


124-17-4 20427 125 31000a minus2819 152E-03 150E-03 4660 80 4181 EDETOX






16721 163 479893 minus2091 811E-03 779E-03 37393








Modelingmethod


n r2






A JJ

JJ

max

max

max

max

( ) =le

lt legt

⎧⎨⎪

⎩⎪

10 0 140 0 1 1080 10

(8)




3 Model validation







Modelingmethod


n r2







































Kpcorr KpKp MW= +

⎛⎝⎜

⎞⎠⎟

= + times⎛⎝⎜

⎞⎠⎟

=

12 6

0 485 10 485 204 36

2 60 13



log S model Result









5 Conclusion


Funding source






Acknowledgment


References
































































Introduction




Calculating Jmax

Calculate Kp

Calculate log Kow




Model validation






Conclusion

Funding source



Acknowledgment

References





23 Calculating Jmax




231 Calculate Kp



(n = 93 r2 = 067 SE = 0794 F = 8405)















K cm hK

K MWpcorr p

p

( ) =+

times1

2 6

(5)









Table 1 (continued)








(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source



54464-57-2 23438 431 60 minus0846 143E-01 775E-02 465 40 1651 RIFM Db





1222-05-5 25841 590a 18a 0079 120E+00 143E-01 025 40 516 RIFM Db









1205-17-0 19222 214 1390 minus1996 101E-02 957E-03 1330 80 501 RIFM Db










31906-04-4 21032 256 6100a minus1896 127E-02 119E-02 7241 80 364 RIFM Db






Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source


18218 219 262843 minus1843 143E-02 133E-02 35088





















Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source


124-17-4 20427 125 31000a minus2819 152E-03 150E-03 4660 80 4181 EDETOX






16721 163 479893 minus2091 811E-03 779E-03 37393








Modelingmethod


n r2






A JJ

JJ

max

max

max

max

( ) =le

lt legt

⎧⎨⎪

⎩⎪

10 0 140 0 1 1080 10

(8)




3 Model validation







Modelingmethod


n r2







































Kpcorr KpKp MW= +

⎛⎝⎜

⎞⎠⎟

= + times⎛⎝⎜

⎞⎠⎟

=

12 6

0 485 10 485 204 36

2 60 13



log S model Result









5 Conclusion


Funding source






Acknowledgment


References
































































Introduction




Calculating Jmax

Calculate Kp

Calculate log Kow




Model validation






Conclusion

Funding source



Acknowledgment

References













K cm hK

K MWpcorr p

p

( ) =+

times1

2 6

(5)









Table 1 (continued)








(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source



54464-57-2 23438 431 60 minus0846 143E-01 775E-02 465 40 1651 RIFM Db





1222-05-5 25841 590a 18a 0079 120E+00 143E-01 025 40 516 RIFM Db









1205-17-0 19222 214 1390 minus1996 101E-02 957E-03 1330 80 501 RIFM Db










31906-04-4 21032 256 6100a minus1896 127E-02 119E-02 7241 80 364 RIFM Db






Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source


18218 219 262843 minus1843 143E-02 133E-02 35088





















Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source


124-17-4 20427 125 31000a minus2819 152E-03 150E-03 4660 80 4181 EDETOX






16721 163 479893 minus2091 811E-03 779E-03 37393








Modelingmethod


n r2






A JJ

JJ

max

max

max

max

( ) =le

lt legt

⎧⎨⎪

⎩⎪

10 0 140 0 1 1080 10

(8)




3 Model validation







Modelingmethod


n r2







































Kpcorr KpKp MW= +

⎛⎝⎜

⎞⎠⎟

= + times⎛⎝⎜

⎞⎠⎟

=

12 6

0 485 10 485 204 36

2 60 13



log S model Result









5 Conclusion


Funding source






Acknowledgment


References
































































Introduction




Calculating Jmax

Calculate Kp

Calculate log Kow




Model validation






Conclusion

Funding source



Acknowledgment

References



(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source



54464-57-2 23438 431 60 minus0846 143E-01 775E-02 465 40 1651 RIFM Db





1222-05-5 25841 590a 18a 0079 120E+00 143E-01 025 40 516 RIFM Db









1205-17-0 19222 214 1390 minus1996 101E-02 957E-03 1330 80 501 RIFM Db










31906-04-4 21032 256 6100a minus1896 127E-02 119E-02 7241 80 364 RIFM Db






Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source


18218 219 262843 minus1843 143E-02 133E-02 35088





















Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source


124-17-4 20427 125 31000a minus2819 152E-03 150E-03 4660 80 4181 EDETOX






16721 163 479893 minus2091 811E-03 779E-03 37393








Modelingmethod


n r2






A JJ

JJ

max

max

max

max

( ) =le

lt legt

⎧⎨⎪

⎩⎪

10 0 140 0 1 1080 10

(8)




3 Model validation







Modelingmethod


n r2







































Kpcorr KpKp MW= +

⎛⎝⎜

⎞⎠⎟

= + times⎛⎝⎜

⎞⎠⎟

=

12 6

0 485 10 485 204 36

2 60 13



log S model Result









5 Conclusion


Funding source






Acknowledgment


References
































































Introduction




Calculating Jmax

Calculate Kp

Calculate log Kow




Model validation






Conclusion

Funding source



Acknowledgment

References

Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source


18218 219 262843 minus1843 143E-02 133E-02 35088





















Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source


124-17-4 20427 125 31000a minus2819 152E-03 150E-03 4660 80 4181 EDETOX






16721 163 479893 minus2091 811E-03 779E-03 37393








Modelingmethod


n r2






A JJ

JJ

max

max

max

max

( ) =le

lt legt

⎧⎨⎪

⎩⎪

10 0 140 0 1 1080 10

(8)




3 Model validation







Modelingmethod


n r2







































Kpcorr KpKp MW= +

⎛⎝⎜

⎞⎠⎟

= + times⎛⎝⎜

⎞⎠⎟

=

12 6

0 485 10 485 204 36

2 60 13



log S model Result









5 Conclusion


Funding source






Acknowledgment


References
































































Introduction




Calculating Jmax

Calculate Kp

Calculate log Kow




Model validation






Conclusion

Funding source



Acknowledgment

References




Table 2 (continued)


(μgcm3)log Kp Kp

(cmh)Kp

corr

(cmh)Jmax


Est Exp Source


124-17-4 20427 125 31000a minus2819 152E-03 150E-03 4660 80 4181 EDETOX






16721 163 479893 minus2091 811E-03 779E-03 37393








Modelingmethod


n r2






A JJ

JJ

max

max

max

max

( ) =le

lt legt

⎧⎨⎪

⎩⎪

10 0 140 0 1 1080 10

(8)




3 Model validation







Modelingmethod


n r2







































Kpcorr KpKp MW= +

⎛⎝⎜

⎞⎠⎟

= + times⎛⎝⎜

⎞⎠⎟

=

12 6

0 485 10 485 204 36

2 60 13



log S model Result









5 Conclusion


Funding source






Acknowledgment


References
































































Introduction




Calculating Jmax

Calculate Kp

Calculate log Kow




Model validation






Conclusion

Funding source



Acknowledgment

References



A JJ

JJ

max

max

max

max

( ) =le

lt legt

⎧⎨⎪

⎩⎪

10 0 140 0 1 1080 10

(8)




3 Model validation







Modelingmethod


n r2







































Kpcorr KpKp MW= +

⎛⎝⎜

⎞⎠⎟

= + times⎛⎝⎜

⎞⎠⎟

=

12 6

0 485 10 485 204 36

2 60 13



log S model Result









5 Conclusion


Funding source






Acknowledgment


References
































































Introduction




Calculating Jmax

Calculate Kp

Calculate log Kow




Model validation






Conclusion

Funding source



Acknowledgment

References
































Kpcorr KpKp MW= +

⎛⎝⎜

⎞⎠⎟

= + times⎛⎝⎜

⎞⎠⎟

=

12 6

0 485 10 485 204 36

2 60 13



log S model Result









5 Conclusion


Funding source






Acknowledgment


References
































































Introduction




Calculating Jmax

Calculate Kp

Calculate log Kow




Model validation






Conclusion

Funding source



Acknowledgment

References







5 Conclusion


Funding source






Acknowledgment


References
































































Introduction




Calculating Jmax

Calculate Kp

Calculate log Kow




Model validation






Conclusion

Funding source



Acknowledgment

References






































Introduction




Calculating Jmax

Calculate Kp

Calculate log Kow




Model validation






Conclusion

Funding source



Acknowledgment

References

an in silico skin absorption model for fragrance...

Documents